For the user to avoid extra manual work Metashape offers a feature for automatic detection of ground points. The automatic classification procedure consists of two steps. At the first step, the point cloud is divided into cells of a certain size. In each cell, the lowest point is detected. Triangulation of these points gives the first approximation of the terrain model. Additionally, at this step, Metashape filters out some noise points to be handled as Low Points class. At the second step new point is added to the ground class, providing that it satisfies two conditions: it lies within a certain distance from the terrain model and that the angle between the terrain model and the line to connect this new point with a point from a ground class is less than a certain angle. The second step is repeated while there still are points to be checked.
Dense Cloud is now (in Metashape 2.0) called Point Cloud.
The information on how to run classification in Metashape you can find in our article - Point Cloud Classification. The current article includes two example sets of parameters for aerial photography of a city district and of a rural area with houses.
First of all lets' recall the parameters of ground point classification in Metashape, please check the dialog window below:
For aerial laser scan data, the ability to use return value during classification process is available. There is also a possibility keep existing ground points during classification process.
Max angle (deg) - determines one of the conditions to be checked while testing a point as a ground one, i.e. sets a limitation for an angle between terrain model and the line to connect the point in question with a point from aground class. For nearly flat terrain it is recommended to use the default value of 15 deg for the parameter. It is reasonable to set a higher value if the terrain contains steep slopes.
Max distance (m) - determines one of the conditions to be checked while testing a point as a ground one, i.e. sets a limitation for a distance between the point in question and terrain model. In fact, this parameter determines the assumption for the maximum variation of the ground elevation at a time.
Cell size (m) - determines the size of the cells for point cloud to be divided into as a preparatory step in ground points classification procedure. Cell size should be indicated with respect to the size of the largest area within the scene that does not contain any ground points, e. g. the building or dense forest.
Erosion radius (m) - determines the indentation (in meters) from unclassified points to create an additional area from the object, it is useful when classifying houses and trees to exclude the remaining "stumps" when building DTM.
Return number - determines which return layer to use during classification (Any Return, First Return or Last Return) for aerial LiDAR data. Usually for laser scans, the first returned is the most significant return and will be associated with the highest feature in the landscape (trees, buildings and so on). It is possible to specify to use First return or Last return. Use the Any Return parameter to perform classification at all return levels.
Rural area with houses and a city district
The example point clouds for aerial photography of a rural area with houses and of a city district are presented below.
We run two tests with different classification parameters. The results of classification procedures are presented in the images below.
Test 1.
Max angle - 4
Max distance (m) - 0.19
Cell size (m) - 50
Erosion radius (m) - 0
Test 2.
Max angle - 12
Max distance (m) - 0.2
Cell size (m) - 30
Erosion radius (m) - 0
Compare:
On the classified point cloud of a rural area with houses, the ground points are colored in brown, while the points above the ground are grey. As presented in the images below low vegetation is better defined in Test 2 (on the right) than in Test 1 (on the left).
On the classified point cloud of a city district, the ground points are colored in brown, other points of the cloud (buildings, vegetation, roads) colored according to the chosen palette. As we can see on the images below buildings were fairly well defined in both Tests (Test 1 is on the left and Test 2 is on the right).
Areas with low vegetation and sown fields
The examples of our data set for testing:
Just want to note that usually in such sets of low vegetation, it is difficult to classify from the earth class. Therefore, low vegetation will mainly fall into the earth-class.
We will use an angle equal to 15 (Max angle = 15) because the territory is flat and does not have sharp changes in height. Max distance and Cell size parameters will be changed.
Test 1.
Max distance (m) - 1
Cell size (m) - 0.5
For this type of data set, we recommend using Classify Points tools (Tools > Point Cloud > Classify Points), in this algorithm, the point cloud will be divided into classes and low vegetation will be classified:
The result from this classification:
Classification of point clouds obtained from satellite images
Satellite image data has a large scale and a large survey area. On such sets, objects (such as houses, trees, and so on) are usually not very treatable due to the scale of the survey. Therefore, for urban areas, the examples of the parameters given above may not be suitable. Below are the results of the classification of such a set:
Test 1:
Max angle - 16
Max distance (m) - 0.2
Cell size (m) - 40
Erosion radius (m) - 0
Test 2:
Max angle - 10
Max distance (m) - 0.5
Cell size (m) - 10
Erosion radius (m) - 0
Test 3:
Max angle - 4
Max distance (m) - 0.19
Cell size (m) - 50
Erosion radius (m) - 0
It may also be convenient to use the selection of cloud points by color as a solution, you can find an example of a workflow in our article: Filtering points based on points colors
Erosion radius
Starting with the version of Metashape 1.8 "Erosion radius" parameter was added in the program. The parameter allows you to eliminate the effect of "stumps" remaining from bushes or houses when building a DTM. The parameter is set in meters and indicates the radius from each point of the point cloud. To avoid using this parameter during classification, use the parameter 0:
Classification results when using different parameters for Erosion radius:
- Erosion radius = 0.04
- Erosion radius = 0.3
- Erosion radius = 1
